The White House is accusing China-linked actors of running what it describes as “industrial-scale” campaigns to extract and replicate American artificial intelligence models, escalating tensions in the global AI race.

In a formal memo to federal agencies, the administration says foreign entities—primarily based in China—are using large networks of proxy accounts and technical workarounds to bypass safeguards and harvest capabilities from advanced U.S. systems.

What is confirmed is that the U.S. government is treating these activities as a form of economic espionage, with plans to share intelligence with domestic AI companies and coordinate defensive measures.

The alleged method centers on “distillation,” a technique that uses outputs from powerful models to train cheaper copies, potentially allowing rivals to accelerate development without equivalent investment.

The move reflects a broader shift in U.S. policy as artificial intelligence becomes a core strategic asset tied to national security and economic leadership.

Chinese officials have rejected the accusations as unfounded, while industry experts note that distinguishing illicit extraction from legitimate model use remains technically complex.

The administration is now exploring enforcement options, including penalties and tighter controls, while Congress considers new legal tools to classify large-scale model extraction as prosecutable industrial espionage.